Kenyan diaspora engineer building coordination infrastructure for East Africa — reducing opacity, enabling coordination, amplifying agency in under-resourced systems.
When 300 Kimi agent swarms show up in Nairobi, the tool layer is already ready.
Agent swarms are useful in proportion to the quality of their domain tools. The 30-server East Africa coordination stack is the structured API layer that makes AI agents actually work for Kenya's institutional landscape — not web-searched guesses, structured tool calls.
| Protocol | Repo | Status |
|---|---|---|
| MCP (Model Context Protocol) | mpesa-mcp | 🟢 PyPI |
| A2A (Agent-to-Agent) | kenya-a2a | 🟢 Live |
| Google ADK | kenya-adk | 🟢 Live |
| Server | Domain | PyPI |
|---|---|---|
| mpesa-mcp | M-PESA mobile payments | ✅ |
| mkopo-mcp | Credit scoring | ✅ |
| bima-mcp | Insurance | ✅ |
| soko-mcp | Commodity markets | ✅ |
| sifa-mcp | Reputation/trust | ✅ |
| remit-mcp | Remittance corridors | ✅ |
| kra-mcp | Tax compliance (≈ TurboTax) | ✅ |
| faida-mcp | Capital markets (≈ Fidelity) | ✅ |
| familia-mcp | Family law & inheritance (≈ LegalZoom) | ✅ |
| diaspora-mcp | Full diaspora lifecycle | ✅ |
| Server | Domain | PyPI |
|---|---|---|
| wapimaji-mcp | Drought & water | ✅ |
| nishati-mcp | Energy / KPLC | ✅ |
| usafiri-mcp | Transport / NTSA | ✅ |
| ardhi-mcp | Land administration | ✅ |
| Server | Domain | PyPI |
|---|---|---|
| afya-mcp | Health / SHA (SHIF) | ✅ |
| afya-ya-akili-mcp | Mental health (≈ BetterHelp) | ✅ |
| elimu-mcp | Education / HELB | ✅ |
| kazi-mcp | Labor market | ✅ |
| haki-ya-kazi-mcp | Labour rights (≈ ACAS) | ✅ |
| kilimo-mcp | Precision agriculture (≈ Climate Corp) | ✅ |
| jumuia-mcp | SACCOs & chamas | ✅ |
| church-mcp | Community/religious institutions | ✅ |
| tafsiri-mcp | Translation (Swahili, Kikuyu, Luo) | ✅ |
| Server | Domain | PyPI |
|---|---|---|
| nyumba-mcp | Housing | ✅ |
| habari-mcp | Civic info / Gazette | ✅ |
| mazingira-mcp | Environment / NEMA | ✅ |
| county-mcp | Kenya 47 counties | ✅ |
| historia-mcp | Historical archives | ✅ |
| Server | Domain | PyPI |
|---|---|---|
| civic-agent-kit | Unified civic SDK | ✅ |
| offline-mcp | Local AI inference (Ollama) | ✅ |
easystocktrader · DarajaAI · ShuleAI · KaziAI · OpenResilience Kenya
🐙 GitHub
🤗 HuggingFace — East Africa datasets
📦 PyPI — MCP servers and tools
📊 Kaggle — Kenya data notebooks
📝 Dev.to — 23+ articles
⚡ Glama — MCP directory
ML Credentials (2020)
Stanford ML Specialization · DeepLearning.AI (Andrew Ng)
Verify ↗
Diaspora
Virginia · Building for 47M+ Swahili speakers
"Diagnose structural absences, not competitive gaps."
M-Pesa wasn't missing from the AI agent ecosystem because engineers hadn't tried. It was missing because the engineer who understood it from the inside wasn't building there yet.
The MCP servers in this portfolio were built at the intersection of three things: what the East African institutional layer needs, what AI coordination infrastructure makes possible, and what I uniquely know how to build.
That intersection is where I work.
Five principles that drive everything here:
-
Build rails, not trains. TCP/IP, railroads, M-Pesa — infrastructure creates ecosystems. Each MCP server is a rail. Others build the trains.
-
Diagnose what's missing. Not "how do I compete?" but "what doesn't exist yet?"
-
Stewardship, not ownership. Data sovereignty: communities should steward their own data. That's why
offline-mcpruns on a Raspberry Pi with no API key. -
Compounding beats intensity. The servers were built incrementally. Each one compounds the next.
-
Timing is strategy. The AI infrastructure window is open now. The Nairobi AI hub moment is real. Build now or build later — but the leverage differs.
"Become who you are, build what is missing, steward what you've been given, and leave more order than chaos behind."
Across the Global South, citizens' health records, land titles, and civic data are increasingly extracted by foreign systems as conditions for receiving aid, services, or connectivity.
This portfolio builds the alternative: AI-powered institutional infrastructure that runs locally, processes data in-country, and gives communities control of their own information.
The offline-mcp server is the sovereignty floor — Ollama-backed inference that runs on a Raspberry Pi with no API key and no data leaving the device.
The sii-stack is the production architecture — n8n + LiteLLM with tri-polar routing (Western / Eastern / Sovereign) so no single government or corporation controls the intelligence layer.
Every MCP server in this stack was built so that a Kenyan community health worker, land administrator, or small business owner can get AI assistance without their data becoming someone else's training set.